JP6937782B2 - Image processing method and device - Google Patents

Description

The present application relates to the field of image processing technology, and more particularly to image processing methods and devices.

With the rapid development of Internet technology, image recognition such as face recognition and authenticated image recognition is being used more and more in network services. However, many images have mesh patterns or watermarks. As a result, the image recognition efficiency is lowered and the difficulty of image recognition is increased.

A main object of the present application is to provide an image processing method and a device.

The present application is implemented by the following technical solutions.

An image processing method is provided. The method is: a step of removing interfering factors from the original image using multiple methods to obtain multiple sample images; segmenting each sample image into multiple sample sub-images based on a given segmentation rule. It includes a step of (dividing); a step of identifying a target sub-image from a sample sub-image having the same attribute; and a step of combining a plurality of target sub-images having different attributes into the target image.

Preferably, the step of identifying the target sub-image from the sample sub-images having the same attributes is: with the step of identifying the mathematical parameters of each sample sub-image; using a clustering algorithm, the same attributes based on the mathematical parameters. A step of dividing the sample sub-image having a Includes the step of identifying the target sub-image and;

Preferably, the step of identifying the mathematical parameters of each sample sub-image is: The sample sub-image uses RGB vectors as the mathematical parameters of the sample sub-image and is based on the RGB information of each pixel in the sample sub-image. It comprises the step of generating the RGB vector for an image.

Preferably, the step of identifying the target sub-image from an image set containing the maximum number of sample sub-images is: said as the target sub-image from the image set containing the maximum number of sample sub-images, acquired after clustering. Includes steps to identify sample sub-images that correspond to center points in the image set.

Preferably, the step of combining a plurality of target sub-images with different attributes into the target image is: based on the position coordinates of each pixel in the target sub-image, the plurality of target sub-images having different attributes into the target image. Includes steps to combine.

An image processing device is provided, said device: with an interference removal unit configured to remove interfering factors from the original image using multiple methods to obtain multiple sample images; a predetermined segmentation rule. With an image segmentation unit configured to segment each sample image into multiple sample subimages based on; and with a targeting unit configured to identify the target subimage from sample subimages with the same attributes. Includes a target combination unit configured to combine multiple target sub-images with different attributes into the target image.

Preferably, the target-specific unit is: with a parameter-specific subsystem configured to identify the mathematical parameters of each sample subimage; the sample having the same attributes based on the mathematical parameters using a clustering algorithm. A set-splitting subsystem in which the sub-images are configured to be split into multiple image sets, each image set containing one or more sample sub-images; the target sub-image from the image set containing the maximum number of sample sub-images. Includes a target-specific subsystem that is configured to identify.

Preferably, the parameter-specific subunit uses an RGB vector as the mathematical parameter of the sample sub-image and uses the RGB vector for the sample sub-image based on the RGB information of each pixel in the sample sub-image. Configured to generate.

Preferably, the target-specific subunit identifies the sample sub-image corresponding to the center point in the image set acquired after clustering as the target sub-image from the image set containing the maximum number of sample sub-images. It is configured to do.

Preferably, the target combination unit is configured to combine the plurality of target sub-images having different attributes with the target image based on the position coordinates of each pixel in the target sub-image.

In the present application, it can be seen that the interfering factors in the original image are first removed using a plurality of methods to obtain a plurality of sample images. Next, each of the plurality of sample images can be segmented into a plurality of sample sub-images based on a predetermined segmentation rule. Next, since the target sub-image can be identified from the sample sub-images having the same attributes, the target sub-image most similar to the real image can be identified from the plurality of sample sub-images having the same attributes. Next, a plurality of target sub-images having different attributes can be combined with the target image. Therefore, the obtained target image can be restored to a real image with high accuracy, and the accuracy of subsequent image recognition can be improved.

FIG. 1 is a schematic flow chart showing an image processing method according to an embodiment of the present application.

FIG. 2 is a schematic diagram showing segmentation of a sample image according to an embodiment of the present application.

FIG. 3 is a schematic diagram showing a procedure for identifying a target sub-image from a sample sub-image having the same attributes according to an example of the embodiment of the present application.

FIG. 4 is a schematic structural diagram showing an image processing device according to an embodiment of the present application.

FIG. 5 is a schematic structural diagram showing an image processing device according to an embodiment of the present application.

Examples of implementation are described in detail here and examples of implementation are shown in the accompanying drawings. When the following description refers to an attached drawing, the same numbers in different attached drawings represent the same or similar elements, unless otherwise specified. The implementations described in the examples of implementation below do not represent all implementations consistent with the present application. On the contrary, they are merely examples of devices and methods described in detail in the appended claims and consistent with some aspects of the present application.

The terms used in the present application are used merely to describe the practice and are not intended to limit the present application. The singular terms "one (a)", "said" and "the" used in this application and the appended claims are plural unless otherwise specified in the context. It is intended to include shapes as well. Also, it should be understood that the term "and / or" as used herein refers to and includes any or all possible combinations of one or more related items.

It should be understood that terms such as "first", "second", "third" can be used herein to describe various types of information, but the information is not limited by term. These terms are used simply to distinguish the same type of information. For example, the first information may be referred to as the second information, and the second information may be referred to as the first information without departing from the scope of the present application. Depending on the context, for example, the word "if" used herein may be "while," "when," or "in response to." determining) ”.

In the related technology, some image processing algorithm or image processing tool such as Photoshop (registered trademark) can be used to remove the mesh pattern or watermark in the original image. However, in such a case, the image obtained by removing the mesh pattern or the watermark cannot usually restore the real image of the original image. As a result, image recognition accuracy will be adversely affected.

FIG. 1 is a schematic flow chart showing an image processing method according to an embodiment of the present application.

With reference to FIG. 1, this image processing method can be applied to terminal devices. Examples of terminal devices include intelligent devices such as smartphones, tablet computers, personal digital assistants (PDAs), and PCs. This image processing method can also be applied to a server and is not limited in the present application. This image processing method can include the following steps.

Step 101: In order to obtain a plurality of sample images, a plurality of methods are used to remove the interfering factor from the original image.

In this embodiment, the original image is usually a recognition target image. The original image can contain one or more interfering factors. The interference factor is usually an interference pattern such as a mesh pattern or a watermark added to the real image.

In this practice, in order to obtain a plurality of sample images, the interfering factors in the original image can be removed by using different interfering factor removing methods provided in the related art. For example, interference factors in the original image can be removed using image processing software such as Photoshop®.

Step 102: Each sample image is segmented into a plurality of sample sub-images based on a predetermined segmentation rule.

After a plurality of sample images are obtained based on step 101, each sample image can be segmented into a plurality of sub-images based on a predetermined segmentation rule. For the sake of simplicity, the sub-image obtained after segmentation is referred to as a sample sub-image.

In this implementation, predetermined segmentation rules can be set by the developer. In the predetermined segmentation rule, the unit used for segmentation may be the size of the sample sub-image or the number of sample sub-images, and is not limited in the present application. For example, a sample image is segmented into 25 sample sub-images according to a predetermined segmentation rule, for example, a sample image is segmented into 25 sample sub-images according to a 5x5 rule. May be good.

In this practice, N sample images can be obtained when the interfering factors in the original image are removed using N different methods in step 101. Further, when each sample image is segmented into M sample sub-images, a total of N × M sample sub-images can be obtained. M and N are natural numbers greater than 1.

Step 103: Identify the target sub-image from the sample sub-images that have the same attributes.

In this practice, each sample sub-image has the corresponding attributes. This attribute can be used to indicate the position information of the sample sub-image in the sample image including the sample sub-image. With reference to FIG. 2, it is assumed that image A is a sample image obtained by removing interference in the original image. The sample image can be segmented into 3 × 3 sample sub-images based on a predetermined segmentation rule. The attributes of the nine sample sub-images are A11, A12, A13, A21, ..., And A33, respectively.

In this implementation, the segmentation rule shown in FIG. 2 is used as an example. In order to obtain N sample images, a total of N × 9 sample sub-images can be obtained when the interfering factors in the original image are removed using N different methods. There are N sample sub-images having one of the attributes A11 to A33. In this step, the target sub-image having the attribute A11 can be identified from the N sample sub-images having the attribute A11, and the target sub-image having the attribute A12 can be identified from the N sample sub-images having the attribute A12. Nine target sub-images having A11 to A33 can be identified in the same way.

Step 104: Combine a plurality of target sub-images with different attributes into the target image.

If the target sub-image is identified from sample sub-images with the same attributes based on step 103, the plurality of target sub-images with different attributes can be combined with the target image. For example, multiple target sub-images can be combined with the target image based on the attributes of each target sub-image, or multiple target sub-images can be combined with the target image based on the position coordinates of each pixel in each target sub-image. It can be combined and is not limited in this application.

For example, the segmentation rule shown in FIG. 2 is used as an example. In this step, nine target sub-images having attributes A11 to A33 can be combined into one target image.

What can be seen from the above description is that the interfering factors in the original image can be first removed by using a plurality of methods to obtain a plurality of sample images. Next, each of the plurality of sample images can be segmented into a plurality of sample sub-images based on a predetermined segmentation rule. Next, the target sub-image can be identified from the sample sub-images having the same attributes, and as a result, the target sub-image most similar to the real image can be identified from the plurality of sample sub-images having the same attributes. Next, a plurality of target sub-images having different attributes can be combined with the target image. Therefore, the acquired target image restores the real image with high accuracy to improve the accuracy of subsequent image recognition.

Although optional, in the embodiment of the present application, referring to FIG. 3, the step of identifying the target sub-image from the sample sub-image having the same attributes can include the following steps.

Step 301: Identify the mathematical parameters of each sample subimage.

In this practice, after the sample image is segmented into multiple sample sub-images, the mathematical parameters of each sample sub-image can be identified for subsequent calculations.

Although optional, in the examples of the present application, an RGB vector can be used as a mathematical parameter of the sample sub-image, and an RGB vector for the sample sub-image can be generated based on the RGB information of each pixel in the sample sub-image. For example, first, RGB information such as the RGB value of each pixel in the sample sub-image can be acquired. Next, an RGB vector can be generated based on the RGB information of each pixel. The RGB vector of the sample sub-image may be {R ₁ , R ₂ , ..., R _k } when the sample sub-image contains K pixels, where the RGB value of the i-th pixel is It is R _i , and the value of i is 1 to K.

Step 302: By using a clustering algorithm, sample sub-images with the same attributes are divided into multiple image sets based on mathematical parameters, where each image set contains one or more sample sub-images.

In this embodiment, a plurality of sample sub-images having the same attributes can be divided into a plurality of image sets based on the mathematical parameters of the sample sub-images by using a clustering algorithm. Examples of the clustering algorithm include a DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm, a K-means clustering algorithm, and the like, and are not limited in the present application.

For example, when the DBSCAN clustering algorithm is used, the scanning radius (eps) and the minimum number of points (minPts) can be specified in advance. Each sample sub-image corresponds to one point in the clustering process. The minimum number of points is the minimum number of sample sub-images that can be included in the image set obtained after division. In the clustering step, the relevant calculations can be made based on the mathematical parameters of the sample subimage. For example, the distance between the RGB vectors of the two sample sub-images can be used as the distance between the two sample sub-images.

Step 303: Identify the target sub-image from the image set that contains the maximum number of sample sub-images.

Based on step 302, after the sample sub-images with the same attributes are divided into multiple image sets, the number of sample sub-images contained in each image set is determined, and then the target sub-image is the sample sub-image. Can be identified from the image set containing the maximum number of.

Although optional, in the embodiment of the present application, the sample sub-image corresponding to the center point in the image set acquired after clustering can be specified as the target sub-image from the image set including the maximum number of sample sub-images.

In this practice, the clustering algorithm is used to identify the target sub-image from the sample sub-images that have the same attributes, ensuring that the identified target sub-image is more similar to the real image.

Corresponding to the implementation of the image processing method, the present application further provides the implementation of an image processing device.

The implementation of the image processing device in the present application can be applied to a terminal device or a server. Implementation of this device can be performed using software, hardware, or both. Software implementation is used as an example. As a logical device, the device is formed by reading the corresponding computer program instructions in the non-volatile memory into the memory by the processor of the terminal device or server where the device is located. From the viewpoint of hardware as shown in FIG. 4, FIG. 4 is a structural diagram showing the hardware of the terminal device or server in which the image processing device in the present application is arranged. In addition to the processor, memory, network interface, and non-volatile memory shown in FIG. 4, the terminal device or server on which the device is located in this embodiment typically includes other hardware based on the actual functionality of the terminal device or server. Can include. Details will not be repeated here.

FIG. 5 is a schematic structural diagram showing an image processing device according to an embodiment of the present application.

Referring to FIG. 5, the image processing device 400 can be applied to the terminal device or server shown in FIG. 4 and includes an interference elimination unit 401, an image segmentation unit 402, a target identification unit 403, and a target combination unit 404. The target-specific subunit 403 can further include a parameter-specific subunit 4031, a set-split subunit 4032, and a target-specific subunit 4033.

The interference removal unit 401 is configured to remove interference factors from the original image using a plurality of methods in order to acquire a plurality of sample images.

The image segmentation unit 402 is configured to segment each sample image into a plurality of sample sub-images based on a predetermined segmentation rule.

The target identification unit 403 is configured to identify the target subimage from the sample subimages having the same attributes.

The target combination unit 404 is configured to combine a plurality of target sub-images having different attributes with the target image.

The parameter-specific subunit 4031 is configured to specify the mathematical parameters of each sample sub-image.

The set division subunit 4032 is configured to use a clustering algorithm to divide a sample sub-image having the same attributes into a plurality of image sets based on mathematical parameters. Each image set contains one or more sample sub-images.

The target identification subunit 4033 is configured to identify the target subimage from an image set that includes the maximum number of sample subimages.

Optionally, the parameter-specific subunit 4031 uses an RGB vector as the mathematical parameter of the sample sub-image and generates an RGB vector for the sample sub-image based on the RGB information of each pixel in the sample sub-image. It is composed of.

Although optional, the targeting subunit 4033 identifies the sample subimage corresponding to the center point in the image set acquired after clustering as the target subimage from the image set containing the maximum number of sample subimages. It is configured as follows.

Optionally, the target combination unit 404 is configured to combine a plurality of target subimages with different attributes with the target image based on the position coordinates of each pixel in the target subimage.

For the implementation process of the functions and roles of the unit in the device, the implementation process of the corresponding step can be referred to. Details are not described here.

Device implementation generally corresponds to method implementation. Therefore, for related explanation, the description of the implementation of the method can be referred to. The implementation of the device described is merely an embodiment. The unit described as a separate part may or may not be physically separated, and the part described as a unit may or may not be a physical unit. They may be located in one place or distributed over multiple network units. Some or all of the modules can be selected according to actual needs to achieve the objectives of the solutions of the present application. Those skilled in the art can understand and implement the implementation of the present invention without any ingenuity.

The above description is merely an example of the implementation of the present application. They do not limit the present application. Any modifications, replacements, or improvements made without departing from the spirit and principles of the present application shall fall within the scope of protection of the present application.

Claims (20)

It ’s a computer implementation method.
A step of identifying a first image to be processed, wherein the first image contains one or more interfering factors.
A step of removing the one or more interfering factors from the first image in order to obtain one or more sample images.
A step of segmenting each sample image of the one or more sample images into a plurality of sample sub-images based on a segmentation rule, where each sample sub-image is associated with an attribute and the attribute is within the sample image. Including the position information of the sample sub-image of the step and
In the step of determining one or more target sub-images from the plurality of sample sub-images, each target sub-image comprises a combination of sample sub-images associated with a common attribute, and each target sub-image comprises a combination of sample sub-images. Steps and steps associated with different attributes
A step of generating a target image that restores a real image of the first image by combining the one or more target sub-images based on the different attributes associated with the one or more target sub-images. ,
A computer mounting method that includes. After the step of determining the one or more target sub-images segments each sample image of the one or more sample images into the plurality of sample sub-images.
A step of determining the mathematical parameters of each sample sub-image among the plurality of sample sub-images, and
A step of dividing the sample sub-image associated with the common attribute into one or more image sets based on the determined mathematical parameters, where each image set is one or more sample sub-images. Steps, including images,
The computer implementation method of claim 1, comprising the step of determining a target subimage from an image set that includes a maximum number of sample subimages. For each sample sub-image, the step of determining the mathematical parameters of each sample sub-image of the one or more sample sub-images
A step of generating an RGB vector for the sample sub-image based on the RGB information of each pixel in the sample sub-image.
The computer mounting method according to claim 2, further comprising a step of specifying the RGB vector as the mathematical parameter of the sample sub-image. The computer implementation of claim 2, wherein the sample sub-images associated with the common attributes are subdivided into one or more image sets based on the determined mathematical parameters using a clustering algorithm. Method. The step of determining the target sub-image from the image set including the maximum number of sample sub-images is obtained after clustering as the target sub-image from the image set including the maximum number of sample sub-images. The computer mounting method according to claim 4, further comprising a step of determining a sample sub-image corresponding to the center point in the image set. The computer implementation method according to claim 1, wherein the one or more interfering factors include at least one of a mesh pattern or a watermark included in the first image. The step of removing the one or more interfering factors from the first image is a step of using the plurality of interfering factor removing methods to obtain the one or more sample images, and the one of the above. Alternatively, each of the plurality of sample images comprises a step, which is acquired based on the corresponding interfering factor removing method among the plurality of interfering factor removing methods.
At least one of the plurality of interfering factor removal methods comprises removing the one or more interfering factors using an image processing software application.
The computer mounting method according to claim 1. The attribute associated with a particular sample sub-image represents a position within the sample image associated with the sample image, and the step of generating the target image is at the position coordinates of each pixel within the target sub-image. based on, comprising the step of combining said one or more target sub-images associated with the different attributes in the target image, the computer implemented method of claim 1. A computer-readable storage medium
Identifying a first image to be processed, wherein the first image contains one or more interfering factors.
To remove the one or more interfering factors from the first image in order to obtain a plurality of sample images.
By segmenting each sample image of the plurality of sample images into a plurality of sample sub-images based on the segmentation rule, each sample sub-image is associated with an attribute, and the attribute is the sample in the sample image. Segmentation, including the position information of the secondary image,
By determining a plurality of target sub-images from the plurality of sample sub-images, each target sub-image comprises a combination of sample sub-images associated with a common attribute, and each target sub-image has different attributes. To be associated, to decide,
Combining the one or more target sub-images to generate a target image that restores the real image of the first image, based on the different attributes associated with the one or more target sub-images. A computer-readable storage medium that stores one or more instructions that can be executed by a computer system to perform an operation. The determination of the plurality of target sub-images is performed after each sample image of the plurality of sample images is segmented into the plurality of sample sub-images.
Determining the mathematical parameters of each sample sub-image of the plurality of sample sub-images
Dividing the sample sub-image associated with a common attribute into a plurality of image sets based on the determined mathematical parameters, each image set containing one or more sample sub-images. To divide and
The computer-readable storage medium of claim 9, wherein the target sub-image is determined from an image set that includes a maximum number of sample sub-images. For each sample sub-image, determining the mathematical parameters of each sample sub-image of the plurality of sample sub-images
Generating an RGB vector for the sample sub-image based on the RGB information of each pixel in the sample sub-image.
The computer-readable storage medium according to claim 10, further comprising specifying the RGB vector as the mathematical parameter of the sample sub-image. 10. The computer-readable memory of claim 10, wherein the sample sub-images associated with a common attribute are divided into one or more image sets based on the determined mathematical parameters using a clustering algorithm. Medium. The determination of the target sub-image from the image set is a sample corresponding to the center point in the image set acquired after clustering as the target sub-image from the image set having the maximum number of sample sub-images. Prepare to determine the secondary image,
The computer-readable storage medium according to claim 12. The computer-readable storage medium of claim 9, wherein the one or more interfering factors include at least one of the mesh patterns or watermarks included in the first image. Removing the one or more interfering factors from the first image is the use of one or more interfering factor removing methods to obtain the one or more sample images. The computer-readable storage medium of claim 9, wherein at least one of the plurality of interfering factor removing methods comprises removing, including using a particular image processing software application. The attribute associated with a particular sample sub-image represents a position within the sample image associated with the sample image, and generating the target image can be attributed to the position coordinates of each pixel within the target sub-image. based on comprises combining the plurality of target sub-images associated with the different attributes in the target image, computer-readable storage medium of claim 9. It ’s a computer implementation system.
With one or more computers
One or more computer memory devices interoperably coupled to the one or more computers, comprising a tangible machine-readable storage medium for storing the one or more instructions, the one or more. When implemented by multiple computers
Identifying a first image to be processed, wherein the first image contains one or more interfering factors.
To remove the one or more interfering factors from the first image in order to obtain a plurality of sample images.
By segmenting each sample image of the plurality of sample images into a plurality of sample sub-images based on the segmentation rule, each sample sub-image is associated with an attribute, and the attribute is the sample in the sample image. Segmentation, including the position information of the secondary image,
By determining one or more target sub-images from the plurality of sample sub-images, each target sub-image comprises a combination of sample sub-images associated with a common attribute, and each target sub-image comprises a combination of sample sub-images. Determining and being associated with different attributes
Combining the one or more target sub-images to generate a target image that restores the real image of the first image, based on the different attributes associated with the one or more target sub-images. A computer-implemented system that performs one or more actions. Determining the one or more target sub-images can be done after each sample image of the plurality of sample images is segmented into the plurality of sample sub-images.
Determining the mathematical parameters of each sample sub-image of the plurality of sample sub-images
Dividing the sample sub-image associated with a common attribute into a plurality of image sets based on the determined mathematical parameters, each image set containing one or more sample sub-images. To divide and
17. The computer implementation system of claim 17, comprising determining a target subimage from an image set that includes a maximum number of sample subimages. For each sample sub-image, determining the mathematical parameters of each sample sub-image of the plurality of sample sub-images
Generating an RGB vector for the sample sub-image based on the RGB information of each pixel in the sample sub-image.
The computer mounting system according to claim 18, wherein the RGB vector is specified as the mathematical parameter of the sample sub-image. 18. The computer-implemented system of claim 18, wherein the sample sub-images associated with a common attribute are subdivided into one or more image sets based on the determined mathematical parameters using a clustering algorithm. ..

Images